Quality Assurance Strategies for Brain State Characterization by MEMRI.

Background

Manganese-enhanced MRI (MEMRI) is a powerful approach for mapping brain-wide neural activity and axonal projections in vivo. Yet standardized computational frameworks for voxel-wise and atlas-based characterization of brain states across large experimental cohorts remain limited. NEW

Method

We present methodological advances for preprocessing and statistical analysis of MEMRI datasets to support scalable, reproducible cohort-level analyses. Quality assurance metrics were developed to evaluate images, cohort-level anatomical alignment, and intensity normalization. Using simulated data, we optimized smoothing, effect-size, and cluster-size thresholds to balance sensitivity and specificity in voxel-wise statistical mapping. We developed 'InVivoSegment' to apply to our new InVivo Atlas for segmentation of MEMRI data and interpretation of brain-wide activity.

Results

Quality assurance analyses established benchmarks for Mn(II)-induced signal- and contrast-to-noise evaluation, precise cohort-level alignment at 100 μm isotropic resolution, and robust intensity normalization. Balanced accuracy and Youden's J statistics were calculated from simulated true positive and noise-only intensities, which defined optimal parameters for smoothing kernel, cluster-size and effect-size thresholds during voxel-wise mapping. Segmentation of simulated data demonstrated reliable transformation of voxel-wise results into regional summaries and identified secondary thresholds that minimize noise-driven artifacts. COMPARISON WITH EXISTING

Methods

Approach to optimize correction parameters for statistical mapping using simulations improves voxel- and segment-wise sensitivity compared to FDR/FWE-based correction procedures.

Conclusions

These methodological advances enable scalable, reproducible, brain-wide quantification of longitudinal changes in MEMRI studies, strengthen mechanistic investigation of brain-state dynamics relevant to human health, and provide broadly applicable tools for other neuroimaging studies. Software is maintained in publicly accessible GitHub repositories.